Chapter 11

Computational Climate ScienceResearch Team

11.1 Members

Hirofumi Tomita (Team Leader)

Yoshiyuki Kajikawa (Senior Scientist)

Shin-ichi Iga (Research Scientist)

Seiya Nishizawa (Research Scientist)

Hisashi Yashiro (Research Scientist)

Sachiho Adachi (Research Scientist)

Yoshiaki Miyamoto (Special Postdoctoral Researcher)

Yousuke Sato (Postdoctoral Researcher)

Tsuyoshi Yamaura (Postdoctoral Researcher)

Ryuji Yoshida (Postdoctoral Researcher)

Hiroaki Miura (Visiting Researcher)

Mizuo Kajino (Visiting Researcher)

Hiroshi Taniguchi (Visiting Researcher)

Tomoko Ohtani (Research Assistant)

Keiko Muraki (Research Assistant)

11.2 Research Activities

Our research team conducts the pioneering research work to lead the future climate simulation. Inorder to enhance the reliability of climate model more, we have aimed to construct a new climatemodel based on the further theoretically physical principles. Conducting such a new model needstremendously large computer resources. Therefore, it is necessary to design the model to pull outthe capability of computers as much as possible. Recent development of supercomputers has aremarkable progress. Hence another numerical technique should be needed under the collaborationof hardware research and software engineering for the effective use on the future HPC, including theK computer and Post K computer.

For the above research purpose and background, our team is cooperating with the computationalscientists in other fields and computer scientists. We enhance the research and development for thefuture climate simulations including effective techniques; we build a next-generation climate model.

88

Chapter 11. Computational Climate Science Research Team

The establishment of the above basic and infrastructure research on the K Computer is stronglyrequired, because this research leads to the post K computer or subsequent ones in the future.

We highlight the following studies in this fiscal year.

1. Construction of a new library for climate study:We have proposed the subject “Estimation of different results by many numerical techniquesand their combination” as a synergetic research to MEXT in 2011 through the discussionwith the Strategic 5 fields (SPIRE). We develop a new library for numerical simulation. Theprogress in development of SCALE is reported. NICAM-DC was imported to SCALE as aglobal dynamical core in this fiscal year. The two landmark papers of the SCALE are reported.

2. Grand challenge run for sub-km horizontal resolution run by global cloud-resolving model:Another outstanding simulation of global model NICAM on the K computer, with super-highresolution (870m), has been done. We analyze the simulation in cooperation with the SPIRE3.We report the further comprehensive analysis of convection properties in the simulation.

3. Disaster prevention research in establishment of COE project:Hyogo-Kobe COE establishment project has accepted 5 subjects in 2012. One of subjects is“the computational research of disaster prevention in the Kansai area”. In this subject, one ofsub-subjects is “Examination of heavy-rainfall event and construction of hazard map”, whichour team is responsible for. The tuning of physical properties focusing on the climatologicalprecipitations and the preliminary result by direct downscaling are reported.

11.3 Research Results and Achievements

11.3.1 Construction of a new library for climate study

We are working on research and development of a library (named SCALE) for numerical models influid dynamical field especially in meteorological field. We examined feasibility of numerical schemeand methods for developing new ones which are suite on massive parallel computers especially the Kcomputer. In order to validate the schemes and test their performance in atmospheric simulations, wehave been developing an atmospheric regional model (named SCALE-RM) as a part of the SCALE li-brary. The SCALE library and the SCALE-RM are currently available as open source software at ourweb site (http://scale.aics.riken.jp/). It is also installed on the K computer and is available for the Kcomputer users as an AICS Software (http://www.aics.riken.jp/en/kcomputer/aics-software.html).In this year, we continued to develop components which are necessary for real atmospheric simula-tions; a boundary turbulence scheme, an urban canopy model, nesting system, and preprocessingtools. We also have improved the library and the model for better performance in both physicaland computational aspects. As a remarkable feature, NICAM-DC ( Nonhydrostatic ICosahedralAtmosphere Model Dynamical Core) was equipped to SCALE library as a global dynamical core.

The validation for the physical performance is an important issue as well as code development.In this fiscal year, we tuned the microphysical scheme, focusing on the one-moment bulk method(Tomita et al.2008). Although this scheme was used also in NICAM through the project SPIRE,it depends on the resolution and phenomena we can see. For this purpose, we conducted the seriesof systematic parameter tuning suitable to Japanese western region using the GSM data as theboundary condition and compared the precipitation with AMeDAS data. Figure 11.1 (a) and (b)shows results of hourly precipitation histogram from two typical parameter sets of the microphysics,which has been used in NICAM experiments (Miyakawa et al. 2014, Miyamoto et al.2013). Afterseveral key parameters were swept, we successfully tuned the parameters as shown in Fig.11.1 (c).

In this fiscal year, two landmark paper for SCALE was published. The first paper describes theproof-of-concept like study according to SCALE policy (Sato et al. 2015[16]). The three microphys-ical schemes, the one-moment bulk, two-moment bulk, and spectral bin schemes were compared bysensitivity experiments in which the other components were fixed in SCALE-RM. Since SCALE istargeting to enable self-model inter-comparison easily, this paper is high significant as a SCALE refer-ence paper. The other paper is about the model description of SCALE-RM dynamical core(Nishizawaet al. 2015[13]). In this paper, we reveals that the influence of the grid aspect ratio of horizontalto vertical grid spacing on turbulence in the planetary boundary layer (PBL) in a large-eddy simu-lation (LES). This paper gives a deep suggestion to meteorological LES. One key point is how thefilter length be configured. It should be based on consideration of the numerical scheme. We also

89

Chapter 11. Computational Climate Science Research Team

Figure 11.1: Results of microphysics tuning.

confirmed necessity of a corrective factor for the grid aspect ratio into the mixing length. As shownin Fig.11.2, these remedy generates the theoretical slope of the energy spectrum; otherwise, spuriousenergy piling appears at high wave numbers.

We investigated also the computational performance of SCALE-RM from the viewpoint of strongscale. Figure 11.3 shows the results of the strong scaling experiments for SCALE-LES. The mosttime-consuming part is the dynamics, and its scaling factor tends to be saturated by decreasing theproblem size. This degradation comes from the increasing ratio of the communication time againstthe computational time. On the other hand, the scaling of physics gives relatively ideal scaling. Inaddition, the I/O part is not a bottleneck. To obtain the faster calculation, we implemented severalchoices both for the temporal and spatial difference schemes. As a result, the longer time step canbe obtained in a certain configuration that is 4th order Runge-Kutta scheme in time and 3rd orderadvection scheme in space.

Figure 11.2: The kinetic energy spectrum.

90

Chapter 11. Computational Climate Science Research Team

Figure 11.3: (a) Strong scale of SCALE-RM. (b) Ratio of communication to calculation.

11.3.2 Grand challenge run for sub-km horizontal resolution run by globalcloud-resolving model

Using the K computer, we have succeeded in conducting the global simulation with the world’shighest resolution, 870 m, which is published in the 2013 fiscal year (Miyamoto et al. 2013). In thefiscal year 2015, an additional analyses to reveal the differences in convection properties in variousatmospheric disturbances has been done. We focused on the differences in convection under fourrepresentative cloudy disturbances: Madden-Julian Oscillation, Tropical Cyclones, Mid-latitudinalLows, and Fronts (Miyamoto et al. 2015[12]). In this fiscal year, we summarized the knowledgethat we have obtained so far, as a review paper (Kajikawa et al. 2016[6]). We conducted furthercomprehensive analysis of the global-mean state and the characteristics of deep convection, to clarifythe difference of the essential change by location and environment. By this paper, this project inour team collaborating with SPIRE was closed once. The subsequent collaboration project leads tothe post-K priority project 4.

11.3.3 Hyogo-Kobe COE establish project

In this fiscal year, the boundary conditions inputted to our regional model SCALE-RE was selected.We decided to use the data from the global warming experiments by MRI-AGCM. Figure 11.4 (a)shows the target region. Figure 11.4 (b) shows the histogram of hourly precipitation of downscal-ing result, compared with AMeDAS in the present climate. The simulation period is from June toSeptember in 10 years. Owing to the successful model tuning described in the previous section, theobservation and model result indicate little difference. Figure 11.4 (c) gives comparison betweenthe present and future climates, regarding to the precipitation intensity. The heavy rainfall eventincreases in the whole Japan area, while the frequency of heavy rainfall is not so changed in theJapanese western area. Figure 11.4 (d) shows an expected maximum rainfall intensity that stochas-tically occurs once twenty years. This result indicates that extreme rainfall occurs in the Kyushuarea, but little change occurs in the Kansai area. However, we should note that this result wasobtained from just one scenario and one GCM model output; we have to regard this result as one ofpossible ensembles. For the more reliable result, we should increase the number of scenario and theintegration period.

11.4 Schedule and Future Plan

In the next year, we will continue to further develop, update and maintain the numerical library forthe K computer (SCALE library). We also try to enhance the performance of each existing scheme.Especially, validation of cumulus parameterization is necessary. At the same time, we will work on

91

Chapter 11. Computational Climate Science Research Team

Figure 11.4: The result of the preciptation property in the Japan/Kansai region by the directdownscaling.

the following three projects in the collaboration with other team in AICS and the scientist in otherinstitute.

1. On the Hyogo/Kobe COE establish project, we will continue the long-term climate simulationby using SCALE-RM to examine the heavy rainfall over Kobe city area. Several MRI-AGCMresults for the future climate is used for increase of scenarios. We will obtain the geographicaldistribution of the frequency of heavy rainfall and evaluate it more precisely. For this purpose,the pseudo global warming method will be also employed.

2. We will also contribute to the CREST, Strategic Basic Research Programs “Innovating ”BigData Assimilation” technology for revolutionizing very-short-range severe weather prediction”to develop the main climatological model in SCALE library. In collaboration with the DataAssimilation team in AICS, we developed a prototype of SCALE-LETKF (Local EnsembleTransform Kalman Filter). For this short range forecast, we will pursue both of the computa-tional and physical performance.

3. We join the POST K Science priority project under the collaboration with JAMSTEC. NICAM-LETKF is one of the target applications. Our team will continue to develop and update thatapplication.

11.5 Publications

Journal Articles

[1] D. Goto et al. “Application of a global nonhydrostatic model with a stretched-grid systemto regional aerosol simulations around Japan”. In: Geoscientific Model Development 8 (2015),pp. 235–259.

[2] H H. Kusaka et al. “Assessment of RCM and urban scenarios uncertainties in the climateprojections for August in the 2050s in Tokyo”. In: Climatic Change ( accepted ) (2016).

[3] S. Iga. “Smooth, seamless, and structured grid generation with flexibility in resolution distri-bution on a sphere based on conformal mapping and the spring dynamics method”. In: Journalof Computational Physics 297 (2015), pp. 381–406.

[4] T. Iguchi et al. “Overview of the development of the Aerosol Loading Interface for Cloudmicrophysics In Simulation (ALICIS)”. In: Progress in Earth and Planetary Science 2(45)(2015).

92

Chapter 11. Computational Climate Science Research Team

[5] Y. Kajikawa et al. “Impact of Tropical Disturbance on the Indian Summer Monsoon OnsetSimulated by a Global Cloud-System-Resolving Model”. In: Scientific Online Letters on theAtmosphere 11 (2015), pp. 80–84.

[6] Y. Kajikawa et al. “Resolution dependence of deep convections in a global simulation fromover 10-kilometer to sub-kilometer grid spacing”. In: Progress in Earth and Planetary Science3(16) (2016).

[7] C. Kodama et al. “A 20-Year Climatology of a NICAM AMIP-Type Simulation”. In: Journalof the Meteorological Society of Japan 93 (2015), pp. 393–424.

[8] J. Leinonen et al. “Performance assessment of a triple-frequency spaceborne cloud-precipitationradar concept using a global cloud-resolving model”. In: Atmospheric Measurement Techniques8 (2015), pp. 3493–3517.

[9] Y. Liu et al. “Modeling study on the transport of summer dust and anthropogenic aerosols overthe Tibetan Plateau”. In: Atmospheric Chemistry and Physics 15 (2015), pp. 12581–12594.

[10] Y. Miyamoto and T. Takemi. “A Triggering Mechanism for Rapid Intensification of TropicalCyclones”. In: Journal of the Atmospheric Sciences 72 (2015), pp. 2666–2681.

[11] Y. Miyamoto et al. “A linear thermal stability analysis of discretized fluid equations”. In:Theoretical and Computational Fluid Dynamics 29 (2015), pp. 155–169.

[12] Y. Miyamoto et al. “Does convection vary in different cloud disturbances?” In: AtmosphericScience Letters 16 (2015), pp. 305–309.

[13] S. Nishizawa et al. “Influence of grid aspect ratio on planetary boundary layer turbulence inlarge-eddy simulations”. In: Geoscientific Model Development 8 (2015), pp. 3393–3419.

[14] S. Nishizawa et al. “Martian dust devil statistics from high-resolution large-eddy simulations”.In: Geophysical Research Letters 43 (2016), pp. 4180–4188.

[15] Y. Sato et al. “Horizontal Distance of Each Cumulus and Cloud Broadening Distance DetermineCloud Cover”. In: Scientific Online Letters on the Atmosphere 11 (2015), pp. 75–79.

[16] Y. Sato et al. “Impacts of cloud microphysics on trade wind cumulus: Which cloud microphysicsprocesses contribute to the diversity in a large eddy simulation?” In: Progress in Earth andPlanetary Science 2(23) (2015).

[17] Y. Sato et al. “Unrealistically pristine air in the Arctic produced by current global scale mod-els”. In: Scientific Reports 6(26561) (2016).

[18] H. G. Takahashi et al. “An Oceanic Impact of the Kuroshio on Surface Air Temperature onthe Pacific Coast of Japan in Summer: Regional H2O Greenhouse Gas Effect”. In: Journal ofClimate 28 (2015), pp. 7128–7144.

[19] B. Wang and Y. Kajikawa. “Reply to Comments on ’Interdecadal Change of the South ChinaSea Summer Monsoon Onset”. In: Journal of Climate 28 (2015), pp. 9036–9039.

[20] T. Yamashita et al. “A numerical study on convection of a condensing CO2 atmosphere underearly Mars like conditions”. In: Journal of the Atmospheric Sciences ( in press ) (2016).

[21] T. Yamaura and Y. Kajikawa. “Decadal change in the boreal summer intraseasonal oscillation”.In: Climate Dynamics ( in press ) (2016).

[22] H. Yashiro et al. “Performance evaluation of throughput-aware framework for ensemble dataassimilation: The case of NICAM-LETKF”. In: Geoscientific Model Development 9 (2016),pp. 2293–2300.

[23] H. Zhao, R. Yoshida, and G.B. Raga. “Impact of the Madden?Julian Oscillation on WesternNorth Pacific Tropical Cyclogenesis Associated with Large-Scale Patterns”. In: Journal ofApplied Meteorology and Climatology 54 (2015), pp. 1413–1429.

Conference Papers

[24] Y. Hisashi et al. “Performance Analysis and Optimization of Nonhydrostatic ICosahedral At-mospheric Model (NICAM) on the K Computer and TSUBAME2.5”. In: Proceedings of thePlatform for Advanced Scientific Computing Conference. PASC ’16. 2016, 3:1–3:8.

93

Chapter 11. Computational Climate Science Research Team

Invited Talks

[25] Y. Kajikawa. Deep Moist Convections in the Sub-kilometer Global Simulation. Tropical Precip-itation Systems Workshop 2015 on Intra-seasonal to seasonal climate variation and prediction,Yokohama, Japan, Sep. 3-4. 2016.

[26] Y. Kajikawa et al. Resolution dependence of deep convections in a global simulation from over10-kilometer to sub-kilometer grid spacing. JpGU Annual meeting 2016: International Session,Chiba, Japan, May 22-26. 2016.

[27] Y. Miyamoto et al. Convection on the globe in the subkilometer global simulation. NTU seminarat the deparment of atmospheric science, National Taiwan University, Taiwan, October 5-6.2015.

[28] S. Nishizawa et al. High resolution Large eddy simulation on Martian planetary boundary layer.JpGU Annual meeting 2016: International Session, Chiba, Japan, May 22-26. 2016.

[29] S. Nishizawa et al. High-resolution modeling and big data analysis at RIKEN AICS. 4th ENESWorkshop on High Performance Computing for Climate and Weather, Toulouse, France, Apr.6-7. 2016.

[30] H. Yashiro. Towards Extreme-Scale Weather and Climate Simulation: The Post K Supercom-puter and Our Challenges. ISC High Performance, 2016 Conference, Frankfurt, Germany, June21. 2016.

Posters and Presentations

[31] S. A. Adachi et al. Application of a Meteorological Large Eddy Simulation Model to Prac-tical Atmospheric Phenomena. 13th Atmospheric Sciences and Application to Air Quality(ASAAQ13), Kobe, Nov. 11. 2015.

[32] S. A. Adachi et al. Performance of SCALE on Downscaling Simulation under the Real At-mospheric Conditions. International WS: Issues in downscaling of climate change projection,Tsukuba, Japan, October 6. 2015.

[33] S.A. Adachi et al. Estimation of climate reproducibility in the summer season in the WesternJapan by SCALE. 2015 annual meeting of meteorological society of Japan, Kyoto, October 29.2015.

[34] Y. Miyamoto et al. Linear stability analysis for discretized system. 2015 annual meeting ofJapan Fluid Mechanics Society,Tokyo, September 26-28. 2015.

[35] Y. Miyamoto et al. Predictability of deep moist atmospheric convection in a sub-kilometer globalsimulation. AOGS 12th Annual Meeting, Singapore, Aug 2-7. 2015.

[36] Y. Miyamoto et al. Predictability of onset of moist convection over the global atmosphere. 2015annual meeting of meteorological society of Japan, Tsukuba, May 21-24. 2015.

[37] Y. Miyamoto et al. The global atmospheric simulation with less than 1km resolution. The 30thTSFD Symposium (Turbulence simulation and flow design - climate change and flow analysis),Tokyo, March 12. 2015.

[38] S. Nishizawa et al. High resolution LES experiment on Martian PBL. Japanese-French modelstudies of planetary atmospheres, Kobe, Japan, May 11-15. 2015.

[39] S. Nishizawa et al. Our model development activity in RIKEN AICS:introduction of SCALEand SCALE-LES. Japanese-French model studies of planetary atmospheres, Kobe, Japan, May11-15. 2015.

[40] S. Nishizawa et al. Research and development of a meteorological simulation model for futuremeteorological simulations. International WS: Issues in downscaling of climate change projec-tion, Tsukuba, Japan, October 6. 2015.

[41] Y. Sato et al. Effect of spreading of shallow cloud and their distance on the cloud cover ratio.JpGU Annual meeting 2016: International Session, Chiba, Japan, May 22-26. 2015.

[42] R. Yoshida et al. Implementation of domain nesting method in SCALE-LES and its evaluation.2015 annual meeting of meteorological society of Japan, Tsukuba, May 21-24. 2015.

94

Chapter 11. Computational Climate Science Research Team

[43] R. Yoshida et al. Super-high resolution experiment by using SCALE-LES. The 2nd seminar formeso-scale meteorology, Kagoshima, June 6. 2015.

95